2-D semi-auto monomodality  nonrigid
registration
Image Fusion Systems Research   cs@imgfsr.com         
In semi-automatic image registration, most work is done by the computer; however, the user is kept in the loop to verify the correctness
of individual registration steps and intervene whenever necessary. Examples of semi-automatic image registration are given below. The
user first selects a pair of corresponding points manually to remove large translation, rotation, and scaling differences between the
images. Using the selected point pair, the program will resample the target image to the geometry of the reference image. This step
makes it possible to reduce the search neighborhood in template matching that follows and improves the registration accuracy.

Next, the program is asked to select a number of corresponding landmarks in the images. After finding the correspondences, the user
is consulted about the correctness of the correspondences. If any of the correspondences is found to be incorrect, the user can remove
them by clicking on them. After this step, the program resamples the target image to the geometry of the reference image. The process
is repeated by increasing the number of landmark correspondences until desired registration accuracy is reached. Semi-automactic
programs are preferred over interactive ones because they require less user interaction.
Fig. 1. (Top row) Two images taken from different viewpoints of Diamond Head, Hawaii. (middle-left) Images registered by the rigid
transformation using two corresponding points in the images. (Middle-right) Images registered by the affine transformation using three
corresponding points in the images. (Bottom row) Images registered using additional corresponding points in the images selected
automatically by the program. Surface spline was used as the transformation to register the images.
Fig. 2. (top) Two aerial images of Honolulu, Hawaii. (bottom) Registration of the images by this software with about 50 automatically
determined corresponding landmarks in the images.
Fig. 3. Registration of images of Koko Head, Hawaii. Initially, three corresponding landmarks were interactively selected from the
images and the images were approximately aligned. Then, the  program automatically selects a few dozen additional
correspondences and registered the images.
To purchase a license for this software, follow this link =>